Kitchen is a Python library for artificial neural network with interface compatible with sklearn (scikit-learn). It is a glue code between Lasagne and sklearn. The library is powered with the Theano allowing fast CPU/GPGPU computing.

The library is attempts to comply with the sklearn coding guidelines, especially:

• Simple sklearn predictor API (instantiate, fit(), predict())
• Handling of random numbers (consistent results between different runs of your code)
• Support for sklearn get_params() and set_params()
• Pipeline compatibility
• Pickle compatibility
• Theano backend for CPU/GPGPU computing, Theano functions compiled on-demand

Dependencies:

• numpy
• Lasagne
• Theano
• sklearn

To install using Python PIP (including all dependencies), use:

pip install git+https://github.com/honzas83/kitchen --process-dependency-links

• Multiclass classification of text using TF-IDF and kitchen (20 newsgroups dataset)